Method and system for machine-based extraction and interpretation of textual information

• US 9,846,692 B2
• Filed: 03/20/2015
• Issued: 12/19/2017
• Est. Priority Date: 02/03/2015
• Status: Expired due to Fees

• Current Assignee: ABBYY Production LLC (ABBYY Software)
• Original Assignee: ABBYY Production LLC (ABBYY Software)
• Inventors: Starostin, Anatoly, Smurov, Ivan, Stepanova, Maria
• Primary Examiner(s): He, Jialong
• Application Number:
  US14/663,868
• Publication Number:
  US 20160224537A1
• Time in Patent Office: 1,005 Days
• CPC Class Codes: G06F 16/367 Ontology
• CPC Class Codes: G06F 40/211 Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
• CPC Class Codes: G06F 40/268 Morphological analysis
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

A method and system of machine-based extraction of information from a text document employ the steps of performing semantic/syntactic analysis for sentences of the document to build semantic-syntactic structures of the sentences; applying production rules to the semantic-syntactic structures to generate a set of logical conclusions about objects of information comprised in the document, wherein the production rules are based on linguistic characteristics and lexical-morphological properties of the semantic-syntactic structures and ontologies of subject matters of the sentences; and using the set of logical conclusions about objects of information comprised in the document to build an ontology-based RDF graph.

Claims

1. A method comprising:

performing, by a processor, semantic/syntactic analysis for sentences of a text document to construct semantic-syntactic structures for the sentences;

applying, by the processor, production rules to the semantic-syntactic structures to generate a set of logical conclusions about information objects in the text document, wherein the production rules are based on linguistic characteristics of the semantic-syntactic structures, lexical-morphological properties of the semantic-syntactic structures, and ontologies of subject matters for the sentences, wherein the production rules comprise interpretation rules and identification rules, wherein the interpretation rules specify fragments to be found in the semantic-syntactic structures, wherein the interpretation rules specify corresponding logical statements that form the set of logical conclusions in response to finding the fragments in the semantic-syntactic structures, and wherein the identification rules are used to merge ones of the information objects;

using, by the processor, the set of logical conclusions about the information objects in the text document to build an ontology-based Resource Definition Framework (RDF) graph; and

storing the ontology-based RDF graph in a storage device and providing searching of the stored ontology-based RDF graph as a knowledge base reference.

2. The method of claim 1, wherein performing the semantic/syntactic analysis further comprises performing lexical-morphological analysis of the text document to construct the semantic-syntactic structures for the sentences.

3. The method of claim 2, further comprising performing rough syntactic analysis of the text document to construct a graph of generalized constituents, wherein the graph of generalized constituents comprises potential lexical meanings of words in the sentences and potential syntactic slots expressing types of relationships between the potential lexical meanings of the words.

4. The method of claim 3, further comprising performing precise syntactic analysis of the graph of generalized constituents, wherein the graph of generalized constituents comprises syntactic-semantic dependency parse trees for the sentences, and wherein performing the precise syntactic analysis comprising identifying ones of the syntactic-semantic dependency parse trees that have highest probabilities of expressing meanings for the sentences.

5. The method of claim 4, further comprising constructing the semantic-syntactic structures which are language independent, wherein the words in the sentences are represented in the semantic-syntactic structures by sematic classes that are language independent, and wherein the semantic classes comprises attributes that express lexical, syntactic, and semantic properties of the words in the sentences.

6. The method of claim 1, wherein applying the production rules further comprises identifying matching ones of the interpretation rules, among the production rules, that do not have object conditions, wherein object conditions indicate types of the information objects assigned to corresponding ones of the interpretation rules for application of the corresponding ones of the interpretation rules to portions of syntactic-semantic dependency parse trees for the sentences, and wherein the logical statements are true for the sentences if the fragments are matched with portions of syntactic-semantic dependency parse trees.

7. The method of claim 6, further comprising adding the matching ones of the interpretation rules to a sorted matching queue.

8. The method of claim 7, further comprising terminating applying the production rules in response to determining that the sorted matching queue is empty.

9. The method of claim 7, further comprising, in response to determining that the sorted matching queue is not empty, selecting a highest priority matching from the matchings in the sorted matching queue to generate the set of logical conclusions from ones of the logical statements corresponding to the highest priority matching, and conditionally adding the set of logical conclusions to a bag of statements, wherein using the set of logical conclusions to build the ontology-based RDF graph comprises converting the bag of statements to the ontology-based RDF graph.

10. The method of claim 9, further comprising searching for new matchings in response to determining that conditionally adding the set of logical conclusions to the bag of statements is successful.

11. The method of claim 9, further comprising declaring the set of logical conclusions invalid and terminating applying the production rules in response to determining that conditionally adding the set of logical conclusions to the bag of statements fails.

12. A system comprising:

a processor; and

a memory storing instructions that, when executed by the processor, cause the processor to;

perform semantic/syntactic analysis for sentences of a text document to construct semantic-syntactic structures for the sentences;

apply production rules to the semantic-syntactic structures to generate a set of logical conclusions about information objects in the text document, wherein the production rules are based on linguistic characteristics of the semantic-syntactic structures, lexical-morphological properties of the semantic-syntactic structures, and ontologies of subject matters for the sentences, wherein the production rules comprise interpretation rules and identification rules, wherein the interpretation rules specify fragments to be found in the semantic-syntactic structures, wherein the interpretation rules specify corresponding logical statements that form the set of logical conclusions in response to finding the fragments in the semantic-syntactic structures, and wherein the identification rules are used to merge ones of the information objects;

use the set of logical conclusions about the information objects in the text document to build a Resource Definition Framework (RDF) graph; and

store the ontology-based RDF graph in a storage device and provide searching of the stored ontology-based RDF graph as a knowledge base reference.

13. The system of claim 12, wherein portions of one or more of the performance of the semantic/syntactic analysis, the application of the production rules, or the use of the set of logical conclusions are performed using the computer language Web Ontology Language (OWL).

14. The system of claim 12, further comprising a database to store the semantic-syntactic structures.

15. A non-transitory computer-readable medium storing instructions that, when executed by a processor, cause the processor to:

perform, by the processor, semantic/syntactic analysis for sentences of a text document to construct semantic-syntactic structures for the sentences;

apply, by the processor, production rules to the semantic-syntactic structures to generate a set of logical conclusions about information objects in the text document, wherein the production rules are based on linguistic characteristics of the semantic-syntactic structures, lexical-morphological properties of the semantic-syntactic structures, and ontologies of subject matters for the sentences, wherein the production rules comprise interpretation rules and identification rules, wherein the interpretation rules specify fragments to be found in the semantic-syntactic structures, wherein the interpretation rules specify corresponding logical statements that form the set of logical conclusions in response to finding the fragments in the semantic-syntactic structures, and wherein the identification rules are used to merge ones of the information objects;

use, by the processor, the set of logical conclusions about the information objects in the text document to build a Resource Definition Framework (RDF) graph; and

store the ontology-based RDF graph in a storage device and provide searching of the stored ontology-based RDF graph as a knowledge base reference.

16. The medium of claim 15, wherein the instructions are further to cause the processor to perform portions of one or more of the performance of the semantic/syntactic analysis, the application of the production rules, or the use of the set of logical conclusions using the computer language Web Ontology Language (OWL).